of June 27, 2018.This information is current as

1 Signaling PathwayβCells via TGF-Enhance the Invasion of Glioma Stem-like Tumor-Associated Microglia/Macrophages

BianZhang, Cheng Qian, You-hong Cui, Xia Zhang and Xiu-wu Yi, Qing-liang Wang, Xue-feng Jiang, Lang Yang, PengYi-fang Ping, Lu Chen, Hua-liang Xiao, Bin Wang, Liang Xian-zong Ye, Sen-lin Xu, Yan-hong Xin, Shi-cang Yu,

ol.1103248http://www.jimmunol.org/content/early/2012/06/03/jimmun

published online 4 June 2012J Immunol 

MaterialSupplementary

8.DC1http://www.jimmunol.org/content/suppl/2012/06/04/jimmunol.110324

        average*  

4 weeks from acceptance to publicationFast Publication! •    

Every submission reviewed by practicing scientistsNo Triage! •    

from submission to initial decisionRapid Reviews! 30 days* •    

Submit online. ?The JIWhy

Subscriptionhttp://jimmunol.org/subscription

is online at: The Journal of ImmunologyInformation about subscribing to

Permissionshttp://www.aai.org/About/Publications/JI/copyright.htmlSubmit copyright permission requests at:

Email Alertshttp://jimmunol.org/alertsReceive free email-alerts when new articles cite this article. Sign up at:

Print ISSN: 0022-1767 Online ISSN: 1550-6606. Immunologists, Inc. All rights reserved.Copyright © 2012 by The American Association of1451 Rockville Pike, Suite 650, Rockville, MD 20852The American Association of Immunologists, Inc.,

is published twice each month byThe Journal of Immunology

by guest on June 27, 2018http://w

ww

.jimm

unol.org/D

ownloaded from

by guest on June 27, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

The Journal of Immunology

Tumor-Associated Microglia/Macrophages Enhance theInvasion of Glioma Stem-like Cells via TGF-b1 SignalingPathway

Xian-zong Ye,*,†,1 Sen-lin Xu,*,†,1 Yan-hong Xin,*,†,1 Shi-cang Yu,*,† Yi-fang Ping,*,† Lu Chen,*,†

Hua-liang Xiao,*,† Bin Wang,*,† Liang Yi,‡ Qing-liang Wang,*,† Xue-feng Jiang,*,† Lang Yang,*,†

Peng Zhang,*,† Cheng Qian,*,† You-hong Cui,*,† Xia Zhang,*,† and Xiu-wu Bian*,†

The invasion of malignant glioma cells into the surrounding normal brain tissues is crucial for causing the poor outcome of this

tumor type. Recent studies suggest that glioma stem-like cells (GSLCs) mediate tumor invasion. However, it is not clear whether

microenvironment factors, such as tumor-associated microglia/macrophages (TAM/Ms), also play important roles in promoting

GSLC invasion. In this study, we found that in primary human gliomas and orthotopical transplanted syngeneic glioma, the number

of TAM/Ms at the invasive front was correlated with the presence of CD133+ GSLCs, and these TAM/Ms produced high levels of

TGF-b1. CD133+ GSLCs isolated from murine transplanted gliomas exhibited higher invasive potential after being cocultured

with TAM/Ms, and the invasiveness was inhibited by neutralization of TGF-b1. We also found that human glioma-derived CD133+

GSLCs became more invasive upon treatment with TGF-b1. In addition, compared with CD1332 committed tumor cells, CD133+

GSLCs expressed higher levels of type II TGF-b receptor (TGFBR2) mRNA and protein, and downregulation of TGFBR2 with

short hairpin RNA inhibited the invasiveness of GSLCs. Mechanism studies revealed that TGF-b1 released by TAM/Ms promoted

the expression of MMP-9 by GSLCs, and TGFBR2 knockdown reduced the invasiveness of these cells in vivo. These results

demonstrate that TAM/Ms enhance the invasiveness of CD133+ GSLCs via the release of TGF-b1, which increases the production

of MMP-9 by GSLCs. Therefore, the TGF-b1 signaling pathway is a potential therapeutic target for limiting the invasiveness of

GSLCs. The Journal of Immunology, 2012, 189: 000–000.

Malignant gliomas are the most common primary braintumors. Despite the improvement in therapy during thepast 5 y, the median survival time for glioblastoma, the

most aggressive glioma, is no more than 14 mo after diagnosis (1).The aggressive invasion by malignant glioma cells into sur-rounding normal brain tissues has increasingly been recognized asan important cause for frequent relapse that leads to high mortality(2). During the past decade, emerging evidence supports the no-tion that cancer stem cells (CSCs) or cancer-initiating cells areresponsible for tumor development including invasion and an-giogenesis, therapy resistance, and recurrence (3–6). Gliomastem-like cells (GSLCs) have been demonstrated to possess highly

invasive activity (7, 8). Besides, the invasive glioma cells presentstem cell characteristics with increasing neurosphere formationability and tumorigenicity (4, 9). However, the effect of micro-environment or niche on GSLC invasion remains unclear.The influence of tumor microenvironment on tumor progression

has been emphasized for many years in almost all kinds of tumor,including glioma (10–12). Tumor-infiltrating leukocytes aremobilized and recruited by tumor-derived factors that contributeto the tumor microenvironment. Tumor-associated macrophages(TAMs), the distinct alternatively activated M2 polarized pop-ulation, are the main components of tumor-infiltrating leukocytes(13–15), which are involved in tumor angiogenesis and invasion(14, 16). In addition to infiltrated leukocytes, resident cells in-cluding tissue macrophages such as microglia in the brain are alsoimportant contributors for tumor progression. Of note, one recentstudy suggests that TAMs, arising from peripheral monocytesrather than resident microglia, are the predominant inflammatorycells to infiltrate human gliomas (17). The biological significanceand possible clinical implications of tumor-associated microglia/macrophages (TAM/Ms) in the marginal area of gliomas are notyet fully understood. Furthermore, the high activity of the TGF-bsignaling pathway in human glioma tissues has been associatedwith a poor prognosis (18). Infiltrating leukocytes are responsiblefor the accumulation of TGF-b1 at the invasive front area of tu-mor, whereas glioma cells are reported to produce TGF-b2 (18–21). Once activated, TGF-b binds the type II TGF-b receptor(TGFBR2). The ligand-bound TGFBR2 is then able efficientlyto trans-activate the type I TGF-b receptor (TGFBR1), whichtransduces intracellular signals through canonical Smad-depen-dent and/or Smad-independent pathways, such as ERK, p38, Rac,and PI3K–Akt pathways (19, 20, 22). TGFBR2 knockdown was

*Key Laboratory of Tumor Immunopathology, Third Military Medical University,Ministry of Education of China, Chongqing 400038, China; †Institute of Pathologyand Southwest Cancer Center, Southwest Hospital, Third Military Medical Univer-sity, Chongqing 400038, China; and ‡Department of Neurosurgery, Daping Hospital,Third Military Medical University, Chongqing 400038, China

1X.-z.Y., S.-l.X., and Y.-h.X. contributed equally to this work.

Received for publication November 17, 2011. Accepted for publication May 3, 2012.

This work was supported by grants from the National Basic Research Program ofChina (973 Program, 2010CB529403) and the National Natural Science Foundationof China (30725035, 30930103, and 30872297).

Address correspondence and reprint requests to Prof. Xiu-wu Bian and Prof. Xia Zhang,Institute of Pathology and Southwest Cancer Center, Southwest Hospital, Third MilitaryMedical University, Gaotanyan 30, Shapingba District, Chongqing 400038, China.E-mail addresses: bianxiuwu@263.net (X.-w.B.) and zhangxia45@yahoo.com (X.Z.)

The online version of this article contains supplemental material.

Abbreviations used in this article: CSC, cancer stem cell; GSLC, glioma stem-like cell;IHC, immunohistochemistry; shRNA, short hairpin RNA; TAM, tumor-associatedmacrophage; TAM/M, tumor-associated microglia/macrophage; TGFBR1, type I TGF-breceptor; TGFBR2, type II TGF-b receptor; TGFBR3, type III TGF-b receptor.

Copyright� 2012 by The American Association of Immunologists, Inc. 0022-1767/12/$16.00

www.jimmunol.org/cgi/doi/10.4049/jimmunol.1103248

Published June 4, 2012, doi:10.4049/jimmunol.1103248 by guest on June 27, 2018

http://ww

w.jim

munol.org/

Dow

nloaded from

shown to inhibit the invasion of glioma cells (23). The con-tributions of TGF-b to cancer development processes, such as cellinvasion, immune suppression, and microenvironment modifica-tion, have been well recognized (1, 19, 20). Moreover, recentstudies showed that TGF-b can increase self-renewal capability ofGSLCs (21, 24). Whether and how the TGF-b signaling pathwaycontributes to other biological properties of GSLCs, such as in-vasion, remain to be explored.In the current study, we aimed to understand the role of TAM/Ms

during GSLC invasion with focus on the TGF-b1 signalingpathway. We found that in both human glioma samples and animalmodels, the distribution of TAM/Ms in the marginal area corre-lated with the location of CD133+ glioma cells. GSLCs exhibitedhigh invasive potential, especially in the presence of TAM/Mswithout direct cell–cell contact. TGFBR2 coexpressed withCD133 in GSLCs was found to be responsible for TAM/Ms-mediated invasion of GSLCs. Furthermore, downregulation ofTGFBR2 decreased the invasion of GSLCs both in vitro andin vivo. Our results indicate that TAM/Ms significantly enhancethe invasive capability of GSLCs through paracrine production ofTGF-b1 and the TGF-b1–TGFBR2 signaling pathway.

Materials and MethodsPatient glioma samples

Surgical glioma specimens were obtained from patients in the Departmentof Neurosurgery, Southwest Hospital and Daping Hospital, Third MilitaryMedical University (Chongqing, China). Written informed consent toperform biological studies was obtained from all of the participants, and thestudy was approved by the Ethics Committee of the Southwest Hospital andDaping Hospital at Third Military Medical University. The tumors wereclassified by at least two pathologists according to the 2007 World HealthOrganization classification of CNS tumors (25).

Preparation of CD133+ and CD1332 cells

Murine glioma cell line GL261 was obtained from the German CancerResearch Center and as a gift from Prof. Zi-ling Wang (Beijing Institute ofTransfusion Medicine, Beijing, China). GL261 and five cases of humanprimary glioma specimens were cultured in DMEM (Life Technologies)containing 10% FBS (Life Technologies). CD133+ or CD1332 populationswere isolated by sorting from GL261 cell line and/or primary specimensusing PE anti-mouse CD133 Ab (eBioscience) and PE anti-human CD133Ab (Miltenyi Biotec, Bergisch-Gladbach, Germany), respectively. Beforeseparation, the percentages of CD133+ cells in the samples were assayedusing the Abs described above and isotype controls. For sorting of CD133+

and CD1332 tumor cells from GL261-derived tumors, single-cell sus-pensions were prepared from GL261-derived tumors as described previ-ously (26). Cells were then incubated for 10 min at 4˚C with rat anti-mouseCD16/CD32 Ab (BD Pharmingen). Subsequently, cells were incubated for30 min at 4˚C with PE anti-mouse CD133 Ab (eBioscience). 7-Amino-actinomycin D (BD Pharmingen) was added (1:10) to discriminate be-tween viable and dead cells. The purity of the cells after sorting wasdetected using flow cytometry. Data acquisition and analysis were per-formed on a BD FACSAria II using CellQuestPro software (BD Bio-sciences). Sorted CD133+ tumor cells were maintained in stem cellmedium, whereas CD1332 tumor cells were cultured in DMEM supple-mented with 10% FBS as previously described (27).

Tumor implantation

Three animal models were used in this study. All the animals were pur-chased from the Laboratory Animal Center, Third Military Medical Uni-versity (Chongqing, China). In the first model, 13 106 GL261 cells or 13106 CD133+ cells were implanted s.c. into syngeneic C57BL/6 mice.Tumor-bearing mice were sacrificed 19 d after transplantation. TAMs andCD133+ or CD1332 tumor cells were analyzed or sorted as describedearlier. In the second model, CD133+ or CD1332 cells sorted from GL261-derived tumors were injected orthotopically with cell numbers of either13 103 or 13 104 per mouse into the brains of 6-wk-old female C57BL/6mice, respectively (n = 5). CD133+ tumor cells (2 3 104) infected withlentiviral stocks were transplanted orthotopically into the brains of 6-wk-old female C57BL/6 mice (n = 7). Mice were maintained up to 4 wk. In thethird model, 1 3 103 CD133+ or 5 3 105 CD1332 tumor cells sorted from

primary glioma specimens were injected orthotopically into the brains of6-wk-old female NOD-SCID mice, respectively (n = 5). At 12–16 wkpostinjection, these NOD-SCID mice were euthanized for analysis. Thebrains of C57BL/6 mice and NOD-SCID mice were collected for furtheranalysis. The animal experiments were approved by the Institutional An-imal Care and Use Committee of the Third Military Medical University.

Immunohistochemistry and immunofluorescence staining

Immunohistochemistry (IHC) was performed on paraffin sections of pri-mary glioma specimens to investigate the distribution of stem cell marker-expressing glioma cells and TAM/Ms in the marginal area. Immunofluo-rescence staining was performed on the frozen sections of orthotopicGL261-derived tumors. Mouse anti-nestin (1:200; Chemicon), rabbit anti-Sox2 (1:200; Abcam, Cambridge, U.K.), and rabbit anti-CD133 (1:200;Abcam) were used as GSLCs markers. Rabbit anti-GFAP (1:100; BosterBiotech, Wuhan, China), rabbit anti-MBP (1:200; Millipore), and mouseanti–b-tubulin III (1:100; Chemicon) were used to identify astrocytic,oligodendrocytic, and neuronal phenotypes, respectively. Mouse anti-human CD68 (1:200; Abcam), rabbit anti-human CD163 (1:100; Abnova,Taipei, Taiwan), and rat anti-mouse CD11b (1:200; eBioscience) as well asrat anti-mouse F4/80 (1:100; eBioscience) were used as TAM/Ms markersfor human and mice, respectively. Rabbit anti-TGFBR2 (1:200; Santa CruzBiotechnology), goat anti-TGFBR2 (1:200; Santa Cruz Biotechnology),and mouse anti–TGF-b1 (1:200; Abcam) were also used. IHC and im-munofluorescence staining was performed as previously described (28, 29).

Preparing TAM/Ms and CD11b+F4/80+ macrophages

The procedures for sorting CD11b+F4/80+ TAM/Ms from GL261-derivedtumors were performed as described earlier. For sorting of macrophagesfrom the spleens of tumor-free mice, C57BL/6 mouse spleens weresheared through a 200-mesh stainless steel sieve, and RBCs were lysed byred cell lysis buffer (Tiangen Biotech, Beijing, China) for 10 min at roomtemperature. Rat anti-mouse CD16/CD32 Ab (BD Pharmingen), FITCanti-mouse CD11b, PE anti-mouse F4/80 Abs (eBioscience), and 7-ami-noactinomycin D (BD Pharmingen) were used in these experiments. Theassay for purity and data acquisition and analysis were described earlier.Isolated splenic macrophages and TAM/Ms were cultured in RPMI 1640(Life Technologies) supplemented with 5% FBS (PAA, Pasching, Aus-tralia) for 18 h as controls. Alternatively, 20 ng/ml IL-4 as well as 100 ng/ml LPS plus 20 ng/ml IFN-g were used to stimulate these two types ofmacrophages, respectively, for 18 h as previously described (30). Thesupernatants were collected and stored at –80˚C for ELISA. The cells werewashed twice and prepared for extraction of total RNA.

Colony formation assay

Plate colony formation assay was used to evaluate the self-renewal ability ofCD133+ tumor cells. Freshly sorted CD133+ tumor cells and their CD1332

counterparts were seeded into 24-well plates at 300 cells per well, re-spectively. All samples were plated in triplicate. During colony growth, theculture medium was replaced every 4 d. After 2 wk, the plates were stainedwith crystal violet (0.2% in 2% ethanol) for 20 min, and the coloniescontaining more than 50 cells were counted.

Coculture of TAM/Ms with either CD133+ or CD1332 tumorcells

Sorted TAM/Ms (23 105 cells per insert) were seeded into upper inserts ofTranswell cell culture chambers (6 wells, 0.4-mm pore size; Falcon,Oxnard, CA) and cocultured with the same number of CD133+ or CD1332

tumor cells (2 3 105 cells per well) in the lower chambers without directcontact. To verify the effect of TGF-b1 in the process of TAM/M-mediatedGSLC invasion, the Ab blocking assay was performed. Briefly, TGF-b1neutralizing Ab (1 mg/ml, mouse monoclonal anti2TGF-b1; Abcam) wasincubated during CD133+ and CD1332 tumor cells coculture with TAM/Ms. CD133+ or CD1332 tumor cells alone were cultured in a 6-well plateas controls. Twenty-four hours from initiation of culture, upper insertscontaining TAM/Ms were discarded. CD133+ or CD1332 tumor cellscocultured with or without TAM/Ms were washed and then used forsubsequent experiments.

To enunciate the signaling pathway responsible for the biological activityof CD133+ tumor cells after coculture with TAMs, several inhibitors wereincubated with CD133+ tumor cells before coculture with TAMs. Theinhibitors were used at concentrations as follows: 10 mM SIS3, a specificinhibitor of Smad3; 10 mM LY294002, a specific inhibitor of PI3K; 10 mMPD98059, a specific inhibitor of MEK; and 50 mM NSC23766, a specificinhibitor of Rac. These inhibitors were purchased from Calbiochem (SanDiego, CA). SB203580 (Sigma-Aldrich Chemical), a specific inhibitor of

2 TGF-b1 CONTROLS GLIOMA STEM-LIKE CELL INVASION

by guest on June 27, 2018http://w

ww

.jimm

unol.org/D

ownloaded from

p38, was used at a concentration of 1 mM. All the inhibitors exceptPD98059 were preincubated with CD133+ tumor cells for 1 h, andPD98059 was preincubated with CD133+ tumor cells for 2 h.

Pretreatment of CD133+ or CD1332 tumor cells with TGF-b1

To define further the invasion-promoting effects of TGF-b1, CD133+ orCD1332 tumor cells (2 3 105 cells per well) sorted from primary gliomaspecimens were seeded into a 6-well plate with or without pretreatmentwith TGF-b1 (5 ng/ml; PeproTech, Rocky Hill, NJ). Twenty-four hoursfrom initiation of culture, CD133+ or CD1332 tumor cells with or withoutpretreatment with TGF-b1 were washed and then used for subsequentexperiments.

Invasion assay

The invasion assay was performed using Transwell cell culture chambers(24 wells, 8-mm pore size; Corning). Briefly, upper inserts were coatedwith 30 ml Matrigel (1:1 dilution; BD Biosciences), and then allowed to setfor 30 min at 37˚C. CD133+ or CD1332 tumor cells were harvested after24 h of coculture with TAM/Ms or 24 h of pretreatment with TGF-b1.Then, 2 3 105 CD133+ tumor cells or CD1332 tumor cells were resus-pended in 200 ml of serum-free DMEM and added to the correspondingupper inserts, respectively. DMEM (600 ml) with 10% FBS was added tothe lower chamber. After 24 h, invaded cells were fixed and stained withcrystal violet (0.2% in 2% ethanol) for 20 min. Cells on the upper side ofthe insert membrane were removed with cotton rods. The invaded cellswere counted at a magnification of 3100 under a light microscope in ninedifferent fields for each insert. The experiments were repeated at least threetimes.

TGFBR2 targeting short hairpin RNA

The short hairpin RNA (shRNA) sequences targeting murine TGFBR2 weredesigned using Invitrogen online BLOCK-iT RNAi Designer (http://www.invitrogen.com/rnai). A previously described nonsilencing sequence(59-TTCTCCGAACGTGTCACGT-39) was used as a control (31). The tar-geting sequences of shRNAs on TGFBR2 used in the experiments were59-CCAACAACATCAACCACAA-39 (shRNA1), 59-GGAGAAGATTCCAG-AAGAT-39 (shRNA2), 59-CCTCACTGCTCACTCCTAA-39 (shRNA3), and59-GCTTCGAACACCATGGAAA-39 (shRNA4). The TGFBR2-targetingshRNA sequences or mock sequences were inserted into the pMagic4.1 ex-pressing vector that expresses eGFP (SunBio, Shanghai, China). Lentiviralstocks were generated by transfecting the pMagic4.1 constructs expressingshRNA targeting TGFBR2 into 293T cells, followed by harvesting after48 h. Efficiency of lentiviral-mediated shRNA knockdown of TGFBR2 incells was determined by Western blot analysis and real-time PCR. Thelentiviral stocks were used to infect CD133+ and CD1332 cells at a multi-plicity of infection of 100. Cells stably expressing TGFBR2-targetingshRNAwere sorted by FACS before they were cocultured with TAM/Ms.

Real-time PCR

Total RNAwas extracted with RNAiso reagent (Takara, Dalian, China). Fivehundred nanograms of RNA from each group was used for RT-PCR witha two-step RT-PCR kit (Takara). Real-time PCR was performed withMaxima SYBR Green/Rox qPCR Master Mix (Fermentas) to analyze thedifferential expressions of TGFBR1, TGFBR2, type III TGF-b receptor(TGFBR3), MMP-2, and MMP-9 in tumor cells and the cytokines levels inmacrophages. The expression of GAPDH mRNA was determined fornormalization. The cycling conditions were 5 min at 95˚C, 40 cycles of 10s at 95˚C, 15 s at 60˚C, 20 s at 72˚C, and 5 s at 84˚C. The sequences of theprimer pairs will be disclosed upon request. Real-time PCR was performedin triplicate including non-template controls.

ELISA

The supernatants of macrophage culture were centrifuged before ELISA.IL-10, IL-12p40 (R&D Systems), and TGF-b1 (eBioscience) were mea-sured using commercial ELISA kits according to the manufacturer’sinstructions. Assay sensitivities were 4 pg/ml for IL-10, 1.8 pg/ml for IL-12p40, and 9 pg/ml for TGF-b1.

Western blot analysis

Western blot was performed as described previously (32). Briefly, theproteins of CD133+ and CD1332 tumor cells were extracted by RIPA LysisBuffer (Beyotime Biotech, Nantong, China) with protease inhibitor PMSF(Beyotime Biotech). Proteins were subjected to 10% SDS-PAGE andtransferred onto polyvinylidene fluoride membranes (Millipore). The mem-

branes were blocked with 5% milk and then incubated with rabbit anti-TGFBR2 (1:1000; Santa Cruz Biotechnology), rabbit anti-GAPDH (1:1500;Cell Signaling Technology), rabbit anti–MMP-9 (1:500; Cell SignalingTechnology), and rabbit anti–b-actin (1:500; Cell Signaling Technology)Abs overnight at 4˚C. Membranes were sufficiently washed and then in-cubated with appropriate HRP-conjugated secondary Abs for 2 h at roomtemperature. Proteins were detected by ECL detection reagent. Expres-sions of TGFBR2 or MMP-9 were normalized against GAPDH or b-actin.

Statistical analysis

All experiments were conducted at least three times. The statistical sig-nificance between testing and control groups was analyzed with SPSS 10.0statistical software. When two groups were compared, the unpaired Stu-dent t test was used. For in vivo study, Kaplan–Meier curves and log-rankanalysis were performed. A p value ,0.05 was considered statisticallysignificant.

ResultsPresence of CD11b+ microglia/macrophages in the marginalarea of glioma correlates with the location of CD133+ gliomacells

Either infiltrated leukocytes and/or their counterparts in residence,particularly macrophages, are implicated in the progression of solidtumors (13). Whether or not these tumor-associated cells haveeffects on the cancer stem-like cells remains elusive. We per-formed immunofluorescence staining on the frozen sections oforthotopic murine GL261-derived tumors to investigate the dis-tribution of stem cell marker-expressing glioma cells and TAM/Ms. As shown in Fig. 1A, a relatively high number of CD11b+

cells was found in the invasive front area of glioma, which wascoincident with a corresponding increment of CD133+ gliomacells; that is, GSLCs (Fig. 1A). Likewise, the relatively lowerdensity of GSLCs was consistent with that of CD11b+ cells in themarginal area (Fig. 1B). As shown in the representative images ofFig. 1A and 1B, when the presence of CD133+ cells and CD11b+

cells within a distance of 100 mm from the invasive edge of gli-oma were taken into account, 20 fields were used to determine therelationship between local distributions of CD11b+ cells andGSLCs. Regression analysis revealed a positive linear correlationbetween these two cell types in the marginal area of glioma (Fig.1C). Of note, most of these CD11b+ cells were also F4/80+ asexamined by immunofluorescence staining (Fig. 1D–G). Flowcytometry analysis of CD11b+ myeloid cells isolated from ectopicmurine GL261-derived tumors also confirmed that they were bothCD45+ (Fig. 1H) and F4/80+ (Fig. 1I). Taken together, our dataindicate that these glioma-associated myeloid cells are macro-phages; that is, TAMs and the existence of CD133+ GSLCs isstrongly correlated with the distribution of TAMs in the marginalarea.

GSLC-associated macrophages produce significant amounts ofTGF-b1

It has been widely reported that TAMs are functionally polarized toalternatively activated M2 macrophage in the tumor microenvi-ronment (14). M2 macrophages are characterized by their highproduction of TGF-b1 and IL-10, two most important cytokinesthat often perform immunomodulatory functions (19, 20, 33).Indeed, as shown in Fig. 2A, a high level of TGF-b1 was detectedin the microenvironment surrounding CD11b+ GSLC-associatedmacrophages in the tumor where CD133+ glioma cells werefound. We then purified CD11b+F4/80+ macrophages by flowcytometry sorting from GL261-derived tumors to a purity of94.6% (Supplemental Fig. 1A). These purified sorted GSLC-associated macrophages or TAMs could adhere to the bottom ofplastic culture wells and presented a typical round or oval shape ofmacrophages (Supplemental Fig. 1B). Moreover, CD11b was also

The Journal of Immunology 3

by guest on June 27, 2018http://w

ww

.jimm

unol.org/D

ownloaded from

observed on the surface of these TAMs under immunofluorescencemicroscopy (Supplemental Fig. 1C). In comparison with classi-cally activated M1 macrophages, cytokine profiles of purifiedTAMs were examined. To generate classically activated M1macrophages, regular macrophages were isolated from the spleensof tumor-free mice and then stimulated with 100 ng/ml LPS and20 ng/ml IFN-g for 18 h. It was reported that IL-4 played animportant role in generating alternatively activated M2 macro-phages (26). We therefore treated TAMs with or without 20 ng/mlIL-4 or 100 ng/ml LPS plus 20 ng/ml IFN-g for 18 h, and thereleased cytokines and their mRNA levels were measured. TAMstreated with IL-4 were high in TGF-b1 compared with TAMstreated with LPS plus IFN-g (Fig. 2B, 2C). Although IL-10 pro-duction from TAMs treated with IL-4 was similar to that fromTAMs treated with LPS plus IFN-g, IL-12p40 production wasmuch higher in TAMs treated with LPS plus IFN-g (Fig. 2D).Other cytokines such as IL-1b, IL-6, and TNF-a, were lower inTAMs treated with IL-4 than in TAMs treated with LPS plus IFN-

g (Supplemental Fig. 1D). These results indicate that TAMs inGL261-derived tumors are polarized to alternatively activated M2macrophages by producing high amounts of TGF-b1, and TAMsare skewed to M1 phenotype in the presence of LPS plus IFN-g byproducing huge amounts of IL-12p40.

Culture and characterization of GSLCs

The CD133+ GSLCs from primary malignant glioma specimens,glioma cell line GL261, and GL261-derived tumors were preparedby FACS sorting with the corresponding surface markers. Theheterogeneous contents of CD133+ cells, ranging from 0.1 to14.6%, were exhibited in five cases of human primary gliomaspecimens (Supplemental Fig. 2A). Approximately 0.67% ofCD133+ cells were detected from the GL261 cell line (Supple-mental Fig. 2B). Notably, CD133+ cells significantly increased inGL261-derived tumors or GL261 CD133+ cell-derived tumors,which were 1.84 or 2.05% of total cells, respectively (Supple-mental Fig. 2B). Case 1 and case 2 of primary glioma specimens

FIGURE 1. Presence of TAM/Ms in the marginal area of glioma correlates with CD133+ glioma cells in murine orthotopically transplanted glioma. (A

and B) The distribution of CD11b+ macrophages (green) and CD133+ glioma cells (red) in the invasive front area of orthotopic GL261 transplanted tumor.

Nuclei were counterstained with Hoechst 33258 (blue). Immunofluorescence staining. (A) Relatively high numbers of both cell types. (B) Relatively low

numbers of the both cell types. (C) Linear regression revealed a positive correlation between the distributions of CD11b+ macrophages and CD133+ glioma

cells in murine glioma model. The relationship between local distributions of CD11b+ cells and GSLCs was determined under 20 fields of confocal

microscopy where CD133+ cells and CD11b+ cells were found within a distance of 100 mm from the invasive edge of glioma. (D–G) Distribution of

CD11b+/F4/80+ macrophages in the murine orthotopic glioma. Representative immunofluorescence images of CD11b+/F4/80+ macrophages are presented.

(D) CD11b (green). (E) F4/80 (red). (F) Nuclei were counterstained with Hoechst 33258 (blue). (G) Double positive for CD11b and F4/80 with Hoechst

33258-stained nuclei is shown in merge. (H) Flow cytometry showed the majority of CD11b+ macrophages coexpressed CD45 (upper right). (I) Percentage

of CD11b+F4/80+ TAMs in GL261-derived tumors. Values represent the mean 6 SD of triplicate determinations, and the data represent one of three

independent experiments.

4 TGF-b1 CONTROLS GLIOMA STEM-LIKE CELL INVASION

by guest on June 27, 2018http://w

ww

.jimm

unol.org/D

ownloaded from

as well as GL261-derived tumors were chosen for sorting CD133+

cells as they contained relatively higher numbers of CD133+ cells.After sorting, the purified cells were evaluated by flow cytometry.The purities of CD133+ cells in case 2 and GL261-derived tumorswere more than 90% (Supplemental Fig. 2C). To establish colonyformation capabilities of CD133+ and CD1332 cells, freshlysorted cells were seeded in 24-well plates at 300 cells per well.After 2-wk culture in DMEM supplemented with 10% FBS,CD133+ cells displayed a higher capability for tumor colonyformation than CD1332 cells (Supplemental Fig. 2D). To examinetumor sphere-forming capabilities of CD133+ and CD1332 cells,freshly sorted cells from case 1, case 2, and GL261-derivedtumors were seeded in serum-free neural stem cell medium.Neurospheres were formed by CD133+ cells from all specimens atday 7 after initiation of culture (Supplemental Fig. 2E). The sortedCD133+ cells expressed abundant nestin and Sox2 and lower levelof GFAP. In contrast, CD1332 cells contained very little nestinand Sox2 and high level of GFAP (Supplemental Fig. 3A). Afterculture in medium containing FBS 10% for 1 wk, CD133+ cellsexhibited multilineage differentiation potential, which was pre-sented by high-level expressions of GFAP, MBP, and b-tubulinIII (Supplemental Fig. 3B). To establish the tumorigenicity ofCD133+ cells, cells isolated from GL261-derived tumors wereorthotopically implanted into C57BL/6, whereas cells isolatedfrom primary glioma specimens were orthotopically implantedinto NOD-SCID mice. Both showed high tumorigenicity. As fewas 1 3 103 CD133+ cells were able to initiate tumor formation. Incontrast, 10-fold higher number of CD1332 cells was needed toform tumor in C57BL/6 mice and 500-fold higher number ofCD1332 cells was required in NOD-SCID mice compared withthat required for CD133+ cells (Supplemental Table I). Theseresults indicated that the CD133+ cells from primary gliomaspecimens or GL261-derived tumors had higher tumorigenicity.

GSLC-associated macrophages enhance the invasivecapability of GSLCs via TGF-b1

After obtaining GSLCs and macrophages from GL261-derivedtumors by FACS sorting, we decided to compare the invasive

ability of CD133+ cells with that of CD1332 cells as shown in Fig.3. In monoculture conditions, CD133+ cells displayed higher in-vasive ability than CD1332 cells (Fig. 3A, d versus e, p , 0.01).

FIGURE 2. TGF-b1 expression in TAMs

derived from murine transplanted glioma. (A)

TGF-b1 cytokine (green) was detected in the

extracellular matrix of CD11b+ macrophages

(red). Nuclei were counterstained with Hoechst

33258 (blue). Immunofluorescence staining.

(B) TGF-b1 production from splenic macro-

phages and TAMs without stimulation (Naive)

or stimulated with IFN-g and LPS (I/L) or IL-

4. (C) TGF-b1 mRNA level in the splenic

macrophages and TAMs without stimulation

(Naive) or stimulated with IFN-g and LPS (I/L)

or IL-4. (D) Production of IL-10 and IL-12p40

from splenic macrophages and TAMs without

stimulation (Naive) or stimulated with IFN-g

and LPS (I/L) or IL-4. Data are the mean6 SD

in triplicate, and one of three independent

experiments is presented.

FIGURE 3. TAMs enhance invasiveness of glioma cells. (A) Comparing

the invasive capabilities of CD133+ and CD1332 cells. The invasive capa-

bility of CD133+ cells was higher than that of CD1332 cells in monoculture

conditions (a versus d, p , 0.01). The invasive capability of CD133+ cells

was increased notably after coculture with TAMs (d versus e, p, 0.01). The

use of TGF-b1 neutralizing Ab in coculture system could significantly de-

crease the invasion of CD133+ cells (e versus f, p , 0.01). In contrast, the

invasive ability of CD1332 cells only slightly decreased after the use of

TGF-b1 neutralizing Ab (b versus c, p . 0.05). (B) Representative images

for numbers of invasive cells in each group. Invaded cells were fixed and

stained with crystal violet. Images were taken under light microscopy

(original magnification 310). Values represent the mean 6 SD of triplicate

determinations, and the data represent one of three independent experiments.

The Journal of Immunology 5

by guest on June 27, 2018http://w

ww

.jimm

unol.org/D

ownloaded from

In coculture system, macrophages sorted from GL261-derived tu-mors were seeded into the upper inserts of a 6-well Transwell andcocultured with the same number of CD133+ cells or CD1332

cells for 24 h. After coculture, GSLC-associated macrophages(i.e., TAMs) in the upper inserts were discarded, and CD133+ cellsand CD1332 cells were assayed for invasion (Fig. 3). Comparingwith CD1332 cells, CD133+ cells displayed notably increasedinvasive ability after coculture with TAMs (Fig. 3A, d versus e,p , 0.01).To determine the role of TGF-b1 in the invasive process, TGF-b1

neutralizing Ab was applied in the cocultures. After the use ofTGF-b1 neutralizing Ab in cocultures, the invasiveness of CD133+

cells was significantly reduced (Fig. 3A, e versus f, p , 0.01). Incontrast, the invasive ability of CD1332 cells only slightly de-creased (Fig. 3A, b versus c, p. 0.05). These results suggested thatTGF-b1 derived from GSLC-associated macrophages was criticalin increasing the invasive potential of CD133+ glioma cells.

Presence of human CD68+ macrophages in the invasive frontarea of human glioma correlates with the location of CD133+

glioma cells

We next addressed whether there was association of microglia ormacrophages with glioma in the primary glioma surgically re-moved from patients. Serial paraffin sections of primary humanglioma specimens were analyzed to study the correlation of CD68+

macrophages and CD133+ glioma cells as shown in Fig. 4. IHCstaining of these sections showed CD133+ glioma cells located atthe invasive front area of glioma (Fig. 4A). CD68+ macrophagesin the marginal area of glioma were accompanied by increasingnumbers of CD133+ glioma cells (Fig. 4B). Furthermore, TGF-b1was detected in the microenvironment containing CD68+ macro-phages (Fig. 4C). We also observed that most CD68+ macrophagescoexpressed CD163, a marker for M2 phenotype (Fig. 4D–F).These results suggest that TGF-b1–expressing TAM/Ms may beassociated with GSLC invasion.To address further whether TGF-b1 could contribute to the in-

vasive capability of CD133+ glioma cells, we added TGF-b1 to theCD133+ cells or CD1332 cells that were isolated from primaryglioma specimens from two patients (Supplemental Fig. 2) and thenexamined their invasive capabilities. Not surprisingly, we found thatCD133+ cells displayed higher invasive ability than CD1332 cellswithout pretreatment with TGF-b1 (Fig. 4G, 4I: case 1: a versus c,p , 0.01; Fig. 4H, 4J: case 2: a versus c, p , 0.01). Invasivecapabilities of CD133+ cells of primary gliomas from two differentpatients were found remarkably increased after preincubation withTGF-b1 for 24 h (Fig. 4G, 4H: c versus d, p , 0.01). These resultsindicate that TGF-b1 can promote GSLC invasiveness in vitro.

MMP-9 but not MMP-2 is the major contributor for GSLCinvasiveness

MMPs such as MMP-2 and MMP-9 are known to be associatedwith glioma invasion, and these enzymes are reportedly induced by

FIGURE 4. Association of CD68+ macrophages with CD133+ GSLCs in

human glioblastoma. (A) A representative image of IHC staining showed the

distribution of CD133+ glioma cells (brown) in the marginal area of human

glioblastoma. (B)A representative image of IHC staining showed a great number

of CD68+ macrophages (brown) located in the marginal area of human glio-

blastoma. (C) High level of TGF-b1 was detected in the microenvironment

surrounding CD68+ macrophages in human glioblastoma. (D and E) Represen-

tative IHC images of CD68+ macrophages (D) that coexpress CD163 (E). The

images were taken under light microscopy. Scale bar, 100 mm. (F) The coex-

pression of CD68 (green) and CD163 (red) Ags on the surface of infiltrating

TAMs in human glioma specimens. Nuclei were counterstained with Hoechst

33258 (blue). Immunofluorescence staining. (G and H) Comparing the invasive

capabilities of CD133+ and CD1332 cells. Comparing with CD1332 cells,

the invasion ability of CD133+ cells from case 1 (G) and case 2 (H) was

slightly increased without pretreatment with TGF-b1 (a versus c in case 1,

p , 0.01; a versus c in case 2, p , 0.05). The invasive capability of

CD133+ cells was remarkably increased after pretreatment with TGF-b1

(c versus d, p , 0.01). (I) Representative images for each group of case 1

are presented. (J) Representative images for each group of case 2 are

presented. Invaded cells were fixed and stained with crystal violet. Images

were taken under light microscopy (original magnification 310). Values

represent the mean 6 SD in triplicate, and the data represent one of three

independent experiments.

6 TGF-b1 CONTROLS GLIOMA STEM-LIKE CELL INVASION

by guest on June 27, 2018http://w

ww

.jimm

unol.org/D

ownloaded from

TGF-b1 (34, 35). To delineate further the invasion promotioneffect of TAM-derived TGF-b1, the expression levels of MMP-2and MMP-9 were determined by real-time PCR. As shown in Fig.5A, the changes of MMP-2 mRNA expression were not significantin CD133+ and CD1332 cells, whereas MMP-9 expression wasfound ∼7-fold higher in CD133+ cells than in CD1332 cells.Furthermore, in the presence of TAMs, additional increase ofMMP-9 mRNAwas detected in the CD133+ GSLCs derived fromorthotopic gliomas. TAMs had no effect on the expression ofMMP-2 mRNA in both CD133+ and CD1332 tumor cells. Neu-tralization of TGF-b1 by TGF-b1 neutralizing Ab resulted ina reduction of MMP-9 mRNA in CD133+ tumor cells (Fig. 5A).

The expressions of MMP-2 and MMP-9 mRNAs were alsoexamined in the glioma specimens from two patients. As shown inFig. 5B, in agreement with the findings from the above-mentionedstudy of murine GL261-derived tumors, MMP-2 expression wasvery low in both CD133+ and CD1332 cells, whereas MMP-9expression was much higher in CD133+ but low in CD1332 gli-oma cells. Furthermore, the exogenous TGF-b1 significantlypromoted MMP-9 expression in CD133+ cells but had less effecton CD1332 tumor cells (Fig. 5B). Notably, exogenous TGF-b1had no significant effect on MMP-2 expressions in both CD133+

and CD1332 tumor cells (Fig. 5B). Previous studies suggest thatSmad and MAPK–p38 pathways as well as others could be theintegrated components in TGF-b1 signaling. Thus, TGF-b1–in-duced MMP-9 expression and invasiveness of CD133+ tumor cellswere examined in the presence of SIS3, an inhibitor for Smad3,

and SB203580, a specific inhibitor of MAPK–p38 pathway. Asshown in Fig. 5C and 5D, the inducible effects of TGF-b1 onMMP-9 expression and invasive capability of CD133+ tumorcells were blocked by these two inhibitors, whereas inhibitors forother signaling pathways such as PD98059 (MEK1/2 inhibitor),LY294002 (PI3K inhibitor), and NSC23766 (Rac inhibitor) hadvery minor or no effect at all (data not shown). These resultssuggest that MMP-9 expression on CD133+ cells mediated byTAM/M-derived TGF-b1 contributes to the high invasive capa-bility of GSLCs, and inhibitors for Smad3 and MAPK–p38pathways would be candidate drugs to treat glioma.

High level of TGFBR2 in CD133+ GSLCs

To explore further the underlying mechanism by which TGF-b1mediated GSLCs invasion, we examined the expression of threeTGF-b1 receptors, TGFBR1, TGFBR2, and TGFBR3. The mRNAlevels of these three receptors were measured by real-time PCR.As shown in Fig. 6A, TGFBR2 expression was found to be muchhigher in CD133+ cells than that in CD1332 cells from murineGL261-derived tumors. Next, we examined the expression ofTGFBR2 in two human primary glioma samples. Similar to what

FIGURE 5. High expression of MMP-9 contributes to invasiveness of

CD133+ glioma cells. (A) The expression levels of MMP-2 and MMP-9

mRNA in CD133+ cells and CD1332 cells sorted from GL261-derived

tumors after coculture with TAMs. (B) The expression levels of MMP-2

and MMP-9 mRNA in CD133+ cells and CD1332 cells sorted from pri-

mary glioma specimens after pretreatment with TGF-b1. (C) MMP-9

mRNA expression (top) and protein level (bottom) in the CD133+ cells

from GL261-derived tumors were reduced in the presence of SIS3 or

SB203580. (D) Either SIS3 or SB203580 significantly reduced invasive-

ness of CD133+ cells from GL261-derived tumors. Quantitative analysis

(top); representative images (bottom). Invaded cells were fixed and stained

with crystal violet; images were taken under light microscopy (original

magnification 310). Values represent the mean 6 SD of triplicate deter-

minations, and the data represent one of three independent experiments.

*p , 0.01.

FIGURE 6. High level of TGFBR2 coexpressed with CD133 in GSLCs.

(A) TGFBR1, TGFBR2, and TGBFR3 mRNA expression levels in CD133+

and CD1332 cells sorted from GL261-derived tumors (GL261);

TGFBR2 mRNA expression levels in CD133+ and CD1332 cells sorted

from human primary glioma specimens (Case 1 and Case 2). (B) TGFBR2

protein expression levels in CD133+ cells and CD1332 cells. (C) The

coexpression of TGFBR2 and CD133 on neurospheres derived from

CD133+ cells (top) and on CD1332 cells (bottom) were detected by im-

munofluorescence staining. Nuclei were counterstained with Hoechst

33258 (blue). Values represent the mean6 SD of triplicate determinations,

and the data represent one of three independent experiments.

The Journal of Immunology 7

by guest on June 27, 2018http://w

ww

.jimm

unol.org/D

ownloaded from

we observed in murine GL261-derived tumors, TGFBR2 expres-sion was higher in human CD133+ cells but not CD1332 cells(Fig. 6A). In parallel to their mRNA levels, higher protein level ofTGFBR2 was detected in the CD133+ cells derived from bothhuman glioma samples and murine GL261-derived tumors thanthat in their CD1332 counterparts (Fig. 6B). Furthermore, im-munofluorescence double staining also confirmed the coex-pression of TGFBR2 and CD133 in the neurospheres derived fromCD133+ cells that were isolated from GL261 and the samplesfrom two patients, respectively, whereas the expressions ofTGFBR2 and CD133 were significantly lower in CD1332 tumorcells (Fig. 6C). Taken together, these data indicate that TGFBR2 isthe key receptor for TGF-b1–mediated GSLC invasion.

Disruption of the TGF-b1 signaling pathway by knockingdown TGFBR2 reduces the invasion capability of GSLCs

To determine whether the invasive process of GSLCs was con-tributed to by the TGFBR2 pathway, four pairs of murine-specificTGFBR2 shRNAs were individually transfected into CD133+ orCD1332 cells that were derived from GL261 tumors to knockdown endogenous TGFBR2. TGFBR2 mRNA in CD133+ tumorcells transfected with lentiviral stocks expressing shRNA1,shRNA2, shRNA3, or shRNA4 were reduced by 61.3, 70.2, 47.9,or 54.6%, respectively (Fig. 7A). Because TGFBR2 protein wasremarkably decreased in the CD133+ tumor cells that weretransduced with shRNA2 (Fig. 7A, inset), we therefore choseshRNA2 for the next experiments. As shown in Fig. 7B, althoughTGFBR2 knockdown in CD1332 cells had no effect on their in-vasion capability, CD133+ tumor cells for which TGFBR2 wasknocked down after transduction of shRNA2 significantly losttheir invasive capability when they were cocultured with TAMs.Loss of invasiveness capability of the CD133+ cells with TGFBR2knockdown was correlated to the significant reduction of MMP-9expression in these cells (Fig. 7C). These CD133+ tumor cellswith TGFBR2 knockdown were then injected orthotopically intothe brains of C57BL/6 mice to examine their capability of growthin situ. Mice that were orthotopically implanted with CD133+

tumor cells with TGFBR2 knockdown exhibited significantlyprolong survival time by 5 to 7 d (Fig. 7D, p, 0.05). The invasionof glioma into surrounding normal brain tissue was examined at3 wk after injection with immunofluorescence microscopy. Asshown in Fig. 7E and 7F, the invasion of glioma derived frominjection was significantly decreased. CD133+ cells were re-markably reduced in the marginal area of transplanting tumorsderived from TGFBR2 knockdown cells (Fig. 7G, mock versusshRNA2, p , 0.01). Thus, our data demonstrate that the reductionof TGFBR2 and disruption of TGF-b1 signaling pathway leads tothe decreased invasive capability of CD133+ tumor cells bothin vitro and in vivo.

DiscussionIn our previous work, we demonstrated a positive correlationbetween the participation of TAM/Ms and the density of GSLCsand confirmed that GSLCs played an important role in the re-cruitment of TAMs (29). In the current study, we examinedwhether TAM/Ms contributed to the invasion of GSLCs. By ex-amining both human primary glioma samples and murine gliomaxenograft, we observed that TAM/Ms were heavily distributed atthe invasive front area of glioma, which was correlated withCD133+ glioma cells (i.e., GSLCs). These TAM/Ms producedhigh amounts of TGF-b1. Moreover, we found that GSLCs hadenhanced invasive ability after coculture with TAM/Ms or pre-treatment with exogenous TGF-b1. Finally, we determined that

FIGURE 7. Knockdown of TGFBR2 by shRNA significantly reduces the

expression of MMP-9 and inhibits the invasion capability of GSLCs. (A)

TGFBR2 mRNA expression in the cells infected with mock, shRNA1,

shRNA2, shRNA3, and shRNA4. All shRNAs remarkably downregulated

TGFBR2 mRNA expressions and its protein level (inset). shRNA2 was the

strongest one, and the other shRNAs had less effect. Values represent the

mean6 SD of triplicate determinations, and the data represent one of three

independent experiments. (B) shRNA2 significantly inhibited the invasive

capability of CD133+ cells but not CD1332 cells. Mock of shRNA exhibited

no effect on the invasive capabilities in both CD133+ and CD1332 cells.

Values represent the mean 6 SD of triplicate determinations, and the data

represent one of three independent experiments. (C) The mRNA expressions

of MMP-9 in CD133+ cells and CD1332 cells infected with mock or

shRNA2 after coculturewith TAMs. shRNA2 significantly inhibitedMMP-9

mRNA expression of CD133+ cells but had no effect on that of CD1332

cells. Values represent the mean 6 SD of triplicate determinations, and the

data represent one of three independent experiments. (D) Knockdown of

TGFBR2 decreased oncogenic potential of GSLCs and increased the sur-

vival of mice bearing intracranial GL261 tumors. Kaplan–Meier curves

demonstrated increased survival time with TGFBR2 knockdown in C57BL/

6 mice injected with 2 3 104 GSLC-infected lentiviral stocks. One repre-

sentative from three independent experiments is shown (seven mice per

group). (E and F) The invasion of CD133+ cells (red) infected with mock and

shRNA2 into surrounding normal brain tissues in vivo. Knockdown of

TGFBR2 inhibited the invasion of glioma cells (eGFP, green), especially the

CD133+ glioma cells (green in plasma and red in cell membrane). Nuclei

were counterstained with Hoechst 33258 (blue). Immunofluorescence

staining. (G) The invasive ability of CD133+ cells infectedwith shRNA2was

significantly decreased in vivo (p, 0.01). Quantitation of invading CD133+

cells was determined under 5 fields of confocal microscopy where the

CD133+/eGFP+ glioma cells were found within a distance of 100 mm from

the invasive edge. Values represent the mean 6 SD of triplicate determi-

nations, and the data represent one of three independent experiments.

8 TGF-b1 CONTROLS GLIOMA STEM-LIKE CELL INVASION

by guest on June 27, 2018http://w

ww

.jimm

unol.org/D

ownloaded from

the elevated production of MMP-9 from CD133+ GSLCs byparacrine TGF-b1 from TAM/Ms via TGFBR2 pathway was thekey contributor to the enhanced invasive potential of GSLCs.The tumor microenvironment is indispensable for tumor initiation

and progression (36). As the major contributing factor to the ac-quisition of core hallmark capabilities (e.g., angiogenesis, invasion,and metastasis) and the creation of a procancer microenvironment,inflammation is considered to be an enabling characteristic ofcancer (12). Among tumor-infiltrating immune cells, TAMs are themajor composition of the inflammation-related microenvironmentin tumor (13, 14). Actually, TAMs infiltrate into the tumor lesioneven in an early stage of tumorigenesis (37). The mechanisms re-sponsible for the infiltration of TAMs into the tumor mass havebeen widely studied (16, 38). Regarding the malignant glioma, boththe resident microglia and the macrophages derived from circulat-ing monocytes constitute the glioma-infiltrating immune cells andare the major participants in glioma progression (39). Microglialcells play a dominant role in the early stage, whereas the macro-phages derived from circulating monocytes are the major participantin the later stage to promote glioma growth (39). Thus, it would bepossible to deter glioma from uncontrolled growth through inter-vening in the recruitment of circulating monocytes into the gliomamass. Moreover, the infiltrating TAM/Ms are considered to be im-munosuppressive (M2) phenotype. GSLCs contribute greatly to theM2 polarization by a variety of secreted factors (40, 41).According to Paget’s “seed and soil” theory, the tumor micro-

environment provides the appropriate “soil” or niche that is re-quired for tumor cells, especially CSCs, to proliferate anddisseminate (12, 42). Therefore, it is important to investigate theimpact of the tumor microenvironment on CSCs (20, 43, 44).Tumor-infiltrating immune cells, such as Foxp3+ regulatoryT cells and TAMs, as well as the condition medium of macro-phages, have been reported to drive the development and growthof CSCs (45–47). The dynamic cross talk between the tumormicroenvironment and CSCs orchestrates tumor initiation andprogression, and the aggressive invasion into surrounding normalbrain tissues is a unique property of malignant glioma. Thequestion raised here is whether TAM/Ms affect the invasion ca-pability of GSLCs. Thus, we examined the distribution of GSLCsin both human primary glioma tissues and murine orthotopicGL261-derived tumors. Abundant GSLCs were found in themarginal area of glioma, and a large amount of TAM/Ms wereclustered in the invasive front area of both primary gliomaspecimens and murine glioma model. TAM/Ms were associatedwith distribution of GSLCs at the invasive front area of glioma,accompanied by high production of TGF-b1. These results arein accord with previous reports that CSCs from many kinds oftumors exhibit high invasive ability and TAMs are recruited to theinvasive site to enhance tumor invasion (7, 43, 48, 49). Thus, wehypothesize that TAM/Ms promote GSLC invasion throughparacrine TGF-b1.In accord with previous reports (13, 40), we found that glioma-

infiltrating TAM/Ms are polarized to M2 phenotype by releasinghigh levels of immunosuppressive cytokines, such as IL-10 andTGF-b1. Glioma had been reported to secret IL-4, an importantregulator of M2 polarization (26, 50), so we further detected theexpression levels of immunosuppressive cytokines in TAMs andsplenic macrophages in the presence of IL-4 or LPS plus IFN-g.We found that TGF-b1 production from TAMs remained un-changed in the presence of LPS plus IFN-g, whereas IL-10 pro-duction from TAMs treated with LPS plus IFN-g was similar tothat from TAMs treated with IL-4. However, TAMs treated withLPS plus IFN-g produced huge amounts of IL-12p40 andexpressed high levels of IL-1b, IL-6, and TNF-a, hallmarks of M1

polarization. Thus, TAMs are able to be skewed toward M1phenotype in the presence of LPS plus IFN-g.Furthermore, we demonstrated that GSLCs possessed high in-

vasive ability, especially after coculture with TAMs that werederived from transplanted tumors or pretreatment with exogenousTGF-b1. Previously, we and others found that GSLCs possessedhigh invasive ability (4, 51). The effect of TAMs to promote tumorinvasion has been reported in the literature (12, 52). TAMs havebeen demonstrated to facilitate the tumor cell invasion by in-creasing a broad array of MMPs that can degrade proteins of theextracellular matrix (37, 53). Direct evidence has been presentedthat MMP-9 derived from bone marrow cells contributes to skincarcinogenesis (54). Additionally, TAMs also produce factors suchas TGF-b1 and IL-10 that may enhance tumor immune suppressionand facilitate tumor progression (40, 53). Several reports illustratethe role of TGF-b1 in maintaining the self-renewal property ofGSLCs (21, 24). Tang and colleagues (55) reported that TGF-b1could induce migration of mesenchymal stem cells. Recent studyshowed that treatment of TGF-b Ab reduced the invasion of gliomacells into surrounding normal brain in animal models (56). Our datashowed that TGF-b1 from TAM/Ms promoted GSLC invasion. Adecreased invasive ability was observed in CD133+ tumor cellswhere TGF-b1 was neutralized. This provided a new insight intothe role of TGF-b1 signaling in GSLC invasion.TGFBR2 expression is higher in GSLCs compared with that

in CD1332 tumor cells. Knockdown of TGFBR2 expression inGSLCs significantly decreases the invasiveness of GSLCs evenwhen they are cocultured with TAMs, indicating that TGFBR2plays a critical role to mediate glioma invasion. MMP-2 andMMP-9 are known to be induced by the activation of TGF-b1signaling (34, 35). We established that MMP-9 but not MMP-2was the major effector responsible for GSLC invasion. Our resultsrevealed that the MMP-9–dependent GSLC invasion was depen-dent on TAM/M-derived TGF-b1. Knockdown of TGFBR2 inCD133+ tumor cells prolonged the survival time of animalstransplanted with these cells. Transplant tumor derived fromCD133+ tumor cells with knockdown of TGFBR2 possesses re-markably decreased invasive ability, indicating the role ofTGFBR2 in the TGF-b1–enhanced glioma invasion.In summary, our study addresses the importance of the tumor

microenvironment in glioma invasion. We demonstrate that TAMsinfiltrating the marginal area of glioma and TGF-b1–TGFBR2pathway are the key contributors for enhancing the invasion ofGSLCs thereby promoting glioma progression. Development oftherapeutic strategies against the TGF-b1–TGFBR2 signalingpathway and exploration of effective means for conversion oftumor-infiltrating macrophages into anti-tumor M1 phenotypemay present new countermeasures to disrupt glioma invasion.

AcknowledgmentsWe thank Wei Sun and Li-ting Wang (Central Laboratory, Third Military

Medical University, Chongqing, China) for technical assistance in laser

confocal scanning microscopy.

DisclosuresThe authors have no financial conflicts of interest.

References1. Van Meir, E. G., C. G. Hadjipanayis, A. D. Norden, H. K. Shu, P. Y. Wen, and

J. J. Olson. 2010. Exciting new advances in neuro-oncology: the avenue to a curefor malignant glioma. CA Cancer J. Clin. 60: 166–193.

2. Zagzag, D., K. Salnikow, L. Chiriboga, H. Yee, L. Lan, M. A. Ali, R. Garcia,S. Demaria, and E. W. Newcomb. 2005. Downregulation of major histocom-patibility complex antigens in invading glioma cells: stealth invasion of thebrain. Lab. Invest. 85: 328–341.

The Journal of Immunology 9

by guest on June 27, 2018http://w

ww

.jimm

unol.org/D

ownloaded from

3. Reya, T., S. J. Morrison, M. F. Clarke, and I. L. Weissman. 2001. Stem cells,cancer, and cancer stem cells. Nature 414: 105–111.

4. Molina, J. R., Y. Hayashi, C. Stephens, and M. M. Georgescu. 2010. Invasiveglioblastoma cells acquire stemness and increased Akt activation. Neoplasia 12:453–463.

5. Bao, S., Q. Wu, R. E. McLendon, Y. Hao, Q. Shi, A. B. Hjelmeland,M. W. Dewhirst, D. D. Bigner, and J. N. Rich. 2006. Glioma stem cells promoteradioresistance by preferential activation of the DNA damage response. Nature444: 756–760.

6. Yao, X. H., Y. F. Ping, J. H. Chen, C. P. Xu, D. L. Chen, R. Zhang, J. M. Wang,and X. W. Bian. 2008. Glioblastoma stem cells produce vascular endothelialgrowth factor by activation of a G-protein coupled formylpeptide receptor FPR.J. Pathol. 215: 369–376.

7. Inoue, A., H. Takahashi, H. Harada, S. Kohno, S. Ohue, K. Kobayashi, H. Yano,J. Tanaka, and T. Ohnishi. 2010. Cancer stem-like cells of glioblastoma char-acteristically express MMP-13 and display highly invasive activity. Int. J. Oncol.37: 1121–1131.

8. Reddy, E. M., S. T. Chettiar, N. Kaur, R. Ganeshkumar, V. Shepal,N. C. Shanbhag, and A. Shiras. 2011. Dlxin-1, a member of MAGE family,inhibits cell proliferation, invasion and tumorigenicity of glioma stem cells.Cancer Gene Ther. 18: 206–218.

9. Yu, S. C., and X. W. Bian. 2009. Enrichment of cancer stem cells based onheterogeneity of invasiveness. Stem Cell Rev. 5: 66–71.

10. Hwang, R. F., T. Moore, T. Arumugam, V. Ramachandran, K. D. Amos, A. Rivera,B. Ji, D. B. Evans, and C. D. Logsdon. 2008. Cancer-associated stromal fibroblastspromote pancreatic tumor progression. Cancer Res. 68: 918–926.

11. Hoelzinger, D. B., T. Demuth, and M. E. Berens. 2007. Autocrine factors thatsustain glioma invasion and paracrine biology in the brain microenvironment. J.Natl. Cancer Inst. 99: 1583–1593.

12. Hanahan, D., and R. A. Weinberg. 2011. Hallmarks of cancer: the next gener-ation. Cell 144: 646–674.

13. Mantovani, A., and A. Sica. 2010. Macrophages, innate immunity and cancer:balance, tolerance, and diversity. Curr. Opin. Immunol. 22: 231–237.

14. Qian, B. Z., and J. W. Pollard. 2010. Macrophage diversity enhances tumorprogression and metastasis. Cell 141: 39–51.

15. Solinas, G., S. Schiarea, M. Liguori, M. Fabbri, S. Pesce, L. Zammataro,F. Pasqualini, M. Nebuloni, C. Chiabrando, A. Mantovani, and P. Allavena. 2010.Tumor-conditionedmacrophages secretemigration-stimulating factor: a newmarkerfor M2-polarization, influencing tumor cell motility. J. Immunol. 185: 642–652.

16. Du, R., K. V. Lu, C. Petritsch, P. Liu, R. Ganss, E. Passegue, H. Song,S. Vandenberg, R. S. Johnson, Z. Werb, and G. Bergers. 2008. HIF1alphainduces the recruitment of bone marrow-derived vascular modulatory cells toregulate tumor angiogenesis and invasion. Cancer Cell 13: 206–220.

17. Parney, I. F., J. S. Waldron, and A. T. Parsa. 2009. Flow cytometry and in vitroanalysis of human glioma-associated macrophages. Laboratory investigation. J.Neurosurg. 110: 572–582.

18. Bruna, A., R. S. Darken, F. Rojo, A. Ocana, S. Penuelas, A. Arias, R. Paris,A. Tortosa, J. Mora, J. Baselga, and J. Seoane. 2007. High TGFbeta-Smad ac-tivity confers poor prognosis in glioma patients and promotes cell proliferationdepending on the methylation of the PDGF-B gene. Cancer Cell 11: 147–160.

19. Massague, J. 2008. TGFbeta in Cancer. Cell 134: 215–230.20. Bierie, B., and H. L. Moses. 2010. Transforming growth factor beta (TGF-beta)

and inflammation in cancer. Cytokine Growth Factor Rev. 21: 49–59.21. Penuelas, S., J. Anido, R. M. Prieto-Sanchez, G. Folch, I. Barba, I. Cuartas,

D. Garcıa-Dorado, M. A. Poca, J. Sahuquillo, J. Baselga, and J. Seoane. 2009.TGF-beta increases glioma-initiating cell self-renewal through the induction ofLIF in human glioblastoma. Cancer Cell 15: 315–327.

22. Zhang, Y. E. 2009. Non-Smad pathways in TGF-beta signaling. Cell Res. 19:128–139.

23. Wesolowska, A., A. Kwiatkowska, L. Slomnicki, M. Dembinski, A. Master,M. Sliwa, K. Franciszkiewicz, S. Chouaib, and B. Kaminska. 2008. Microglia-derived TGF-beta as an important regulator of glioblastoma invasion—an inhi-bition of TGF-beta-dependent effects by shRNA against human TGF-beta type IIreceptor. Oncogene 27: 918–930.

24. Ikushima, H., T. Todo, Y. Ino, M. Takahashi, K. Miyazawa, and K. Miyazono.2009. Autocrine TGF-beta signaling maintains tumorigenicity of glioma-initiating cells through Sry-related HMG-box factors. Cell Stem Cell 5: 504–514.

25. Louis, D. N., H. Ohgaki, O. D. Wiestler, W. K. Cavenee, P. C. Burger, A. Jouvet,B. W. Scheithauer, and P. Kleihues. 2007. The 2007 WHO classification oftumours of the central nervous system. Acta Neuropathol. 114: 97–109.

26. DeNardo, D. G., J. B. Barreto, P. Andreu, L. Vasquez, D. Tawfik, N. Kolhatkar,and L. M. Coussens. 2009. CD4(+) T cells regulate pulmonary metastasis ofmammary carcinomas by enhancing protumor properties of macrophages.Cancer Cell 16: 91–102.

27. Yi, L., Z. H. Zhou, Y. F. Ping, J. H. Chen, X. H. Yao, H. Feng, J. Y. Lu,J. M. Wang, and X. W. Bian. 2007. Isolation and characterization of stem cell-like precursor cells from primary human anaplastic oligoastrocytoma. Mod.Pathol. 20: 1061–1068.

28. Ping, Y. F., X. H. Yao, J. Y. Jiang, L. T. Zhao, S. C. Yu, T. Jiang, M. C. Lin,J. H. Chen, B. Wang, R. Zhang, et al. 2011. The chemokine CXCL12 and itsreceptor CXCR4 promote glioma stem cell-mediated VEGF production andtumour angiogenesis via PI3K/AKT signalling. J. Pathol. 224: 344–354.

29. Yi, L., H. Xiao, M. Xu, X. Ye, J. Hu, F. Li, M. Li, C. Luo, S. Yu, X. Bian, andH. Feng. 2011. Glioma-initiating cells: a predominant role in microglia/macrophages tropism to glioma. J. Neuroimmunol. 232: 75–82.

30. Tjiu, J. W., J. S. Chen, C. T. Shun, S. J. Lin, Y. H. Liao, C. Y. Chu, T. F. Tsai,H. C. Chiu, Y. S. Dai, H. Inoue, et al. 2009. Tumor-associated macrophage-induced invasion and angiogenesis of human basal cell carcinoma cells bycyclooxygenase-2 induction. J. Invest. Dermatol. 129: 1016–1025.

31. Sowinski, S., C. Jolly, O. Berninghausen, M. A. Purbhoo, A. Chauveau,K. Kohler, S. Oddos, P. Eissmann, F. M. Brodsky, C. Hopkins, et al. 2008.Membrane nanotubes physically connect T cells over long distances presentinga novel route for HIV-1 transmission. Nat. Cell Biol. 10: 211–219.

32. Yang, L., Y. F. Ping, X. Yu, F. Qian, Z. J. Guo, C. Qian, Y. H. Cui, andX. W. Bian. 2011. Gastric cancer stem-like cells possess higher capability ofinvasion and metastasis in association with a mesenchymal transition phenotype.Cancer Lett. 310: 46–52.

33. Mosser, D. M., and X. Zhang. 2008. Interleukin-10: new perspectives on an oldcytokine. Immunol. Rev. 226: 205–218.

34. Kim, E. S., M. S. Kim, and A. Moon. 2004. TGF-beta-induced upregulation ofMMP-2 and MMP-9 depends on p38 MAPK, but not ERK signaling in MCF10Ahuman breast epithelial cells. Int. J. Oncol. 25: 1375–1382.

35. Chen, D. L., Y. F. Ping, S. C. Yu, J. H. Chen, X. H. Yao, X. F. Jiang, H. R. Zhang,Q. L. Wang, and X. W. Bian. 2009. Downregulating FPR restrains xenografttumors by impairing the angiogenic potential and invasive capability of malig-nant glioma cells. Biochem. Biophys. Res. Commun. 381: 448–452.

36. Albini, A., and M. B. Sporn. 2007. The tumour microenvironment as a target forchemoprevention. Nat. Rev. Cancer 7: 139–147.

37. Lewis, C. E., and J. W. Pollard. 2006. Distinct role of macrophages in differenttumor microenvironments. Cancer Res. 66: 605–612.

38. Murdoch, C., A. Giannoudis, and C. E. Lewis. 2004. Mechanisms regulating therecruitment of macrophages into hypoxic areas of tumors and other ischemictissues. Blood 104: 2224–2234.

39. Gabrusiewicz, K., A. Ellert-Miklaszewska, M. Lipko, M. Sielska, M. Frankowska,and B. Kaminska. 2011. Characteristics of the alternative phenotype of microglia/macrophages and its modulation in experimental gliomas. PLoS ONE 6: e23902.

40. Wu, A., J. Wei, L. Y. Kong, Y. Wang, W. Priebe, W. Qiao, R. Sawaya, andA. B. Heimberger. 2010. Glioma cancer stem cells induce immunosuppressivemacrophages/microglia. Neuro-oncol. 12: 1113–1125.

41. Wei, J., A. Wu, L. Y. Kong, Y. Wang, G. Fuller, I. Fokt, G. Melillo, W. Priebe,and A. B. Heimberger. 2011. Hypoxia potentiates glioma-mediated immuno-suppression. PLoS ONE 6: e16195.

42. Psaila, B., and D. Lyden. 2009. The metastatic niche: adapting the foreign soil.Nat. Rev. Cancer 9: 285–293.

43. Moriyama, T., K. Ohuchida, K. Mizumoto, L. Cui, N. Ikenaga, N. Sato, andM. Tanaka. 2010. Enhanced cell migration and invasion of CD133+ pancreaticcancer cells cocultured with pancreatic stromal cells. Cancer 116: 3357–3368.

44. Ye, X. Z., S. C. Yu, and X. W. Bian. 2010. Contribution of myeloid-derivedsuppressor cells to tumor-induced immune suppression, angiogenesis, invasionand metastasis. J. Genet. Genomics 37: 423–430.

45. Yang, S., B. Wang, C. Guan, B. Wu, C. Cai, M. Wang, B. Zhang, T. Liu, andP. Yang. 2011. Foxp3+IL-17+ T cells promote development of cancer-initiatingcells in colorectal cancer. J. Leukoc. Biol. 89: 85–91.

46. Cao, J. X., Y. X. Cui, Z. J. Long, Z. M. Dai, J. Y. Lin, Y. Liang, F. M. Zheng,Y. X. Zeng, and Q. Liu. 2010. Pluripotency-associated genes in human naso-pharyngeal carcinoma CNE-2 cells are reactivated by a unique epigenetic sub-microenvironment. BMC Cancer 10: 68.

47. Okuda, H., A. Kobayashi, B. Xia, M. Watabe, S. K. Pai, S. Hirota, F. Xing,W. Liu, P. R. Pandey, K. Fukuda, et al. 2012. Hyaluronan synthase HAS2 pro-motes tumor progression in bone by stimulating the interaction of breast cancerstem-like cells with macrophages and stromal cells. Cancer Res. 72: 537–547.

48. Pollard, J. W. 2008. Macrophages define the invasive microenvironment in breastcancer. J. Leukoc. Biol. 84: 623–630.

49. Korkaya, H., A. Paulson, F. Iovino, and M. S. Wicha. 2008. HER2 regulates themammary stem/progenitor cell population driving tumorigenesis and invasion.Oncogene 27: 6120–6130.

50. Ghosh, A., M. Bhattacharya, P. Sarkar, S. Acharya, and S. Chaudhuri. 2010. T11target structure exerts effector function by activating immune cells in CNSagainst glioma where cytokine modulation provide favorable microenvironment.Indian J. Exp. Biol. 48: 879–888.

51. Yu, S. C., Y. F. Ping, L. Yi, Z. H. Zhou, J. H. Chen, X. H. Yao, L. Gao,J. M. Wang, and X. W. Bian. 2008. Isolation and characterization of cancer stemcells from a human glioblastoma cell line U87. Cancer Lett. 265: 124–134.

52. Gocheva, V., H. W. Wang, B. B. Gadea, T. Shree, K. E. Hunter, A. L. Garfall,T. Berman, and J. A. Joyce. 2010. IL-4 induces cathepsin protease activity intumor-associated macrophages to promote cancer growth and invasion. GenesDev. 24: 241–255.

53. Mantovani, A., T. Schioppa, C. Porta, P. Allavena, and A. Sica. 2006. Role oftumor-associated macrophages in tumor progression and invasion. Cancer Me-tastasis Rev. 25: 315–322.

54. Coussens, L. M., C. L. Tinkle, D. Hanahan, and Z. Werb. 2000. MMP-9 suppliedby bone marrow-derived cells contributes to skin carcinogenesis. Cell 103: 481–490.

55. Tang, Y., X. Wu, W. Lei, L. Pang, C. Wan, Z. Shi, L. Zhao, T. R. Nagy, X. Peng,J. Hu, et al. 2009. TGF-beta1-induced migration of bone mesenchymal stemcells couples bone resorption with formation. Nat. Med. 15: 757–765.

56. Hulper, P., W. Schulz-Schaeffer, C. Dullin, P. Hoffmann, J. Harper, L. Kurtzberg,S. Lonning, W. Kugler, M. Lakomek, and B. Erdlenbruch. 2011. Tumor local-ization of an anti-TGF-b antibody and its effects on gliomas. Int. J. Oncol. 38:51–59.

10 TGF-b1 CONTROLS GLIOMA STEM-LIKE CELL INVASION

by guest on June 27, 2018http://w

ww

.jimm

unol.org/D

ownloaded from